Amyloid-β probes: Review of structure–activity and brain-kinetics relationships

Introduction

Alzheimer’s disease (AD) is a progressive neurodegenerative disorder of the central nervous system currently affecting ~5.4 million Americans, a number that could increase to 11–16 million by the year 2050. In the United States, AD represents the 6th leading cause of death. Between 2000 and 2008, the number of deaths caused by AD increased by 66%, a dramatic rise, especially when compared to other causes of death, such as heart disease, stroke, prostate and breast cancer, and HIV, which decreased by 3–29% during that time period [1]. As these numbers indicate, AD represents a significant and increasing burden on our population, and efforts towards the development of new and improved diagnostics and therapeutics for this devastating disease are important research endeavors.

Several pathological hallmarks of AD have been identified, and they include decreased cholinergic neurons and acetylcholine (ACh) levels, plaques caused by aggregation of the protein fragment amyloid-β (Aβ), tangles associated with irregular phosphorylation of tau protein, inflammation and increased oxidative stress from reactive oxygen species (ROS), as well as dyshomeostasis and miscompartmentalization of metal ions such as Cu, Fe, and Zn. Observations of these hallmarks have led to several hypotheses in attempts to explain the underlying cause of the disease, which is likely multifactorial. However, the exact cause of AD still remains unknown.

Postmortem histopathological examination of Aβ plaques is currently the only way to firmly confirm AD [2]. In view of the limited accessibility to living brain and other central nervous system (CNS) tissues, AD is currently diagnosed through memory tests and/or based on the patients’ history [2]. Obviously these kinds of diagnostic tools lack absolute sensitivity and accuracy, especially in the early stages of the disease. Therefore, as Aβ plaques precede the onset of dementia and cognitive decline in AD patients, their detection by nuclear imaging techniques such as positron emission tomography (PET) or single-photon emission computed tomography (SPECT) represents the presymptomatic diagnostic tool of choice for AD [3-5].

The presence of different binding sites in Aβ aggregates led medicinal chemists to investigate and develop a variety of chemical scaffolds as Aβ-imaging tracers [6-9]. To provide a high readable signal-to-background ratio, the ideal Aβ-imaging probes should have certain brain kinetics: a rapid initial brain uptake and a fast washout. Early efforts towards developing Aβ stains focused on dyes such as congo red (1), chrysamine G (2), pinacyanol (3), and thioflavin-T (4) (Figure 1A). However, the bulky and ionic natures of these dyes prevented them from crossing the blood brain barrier (BBB), and consequently, no in vivo benefits were obtained from these initial investigations [10,11]. During the past decade, efforts directed at developing probes that display uptake and retention that differ in healthy and AD-affected brains resulted in a variety of radiolabeled molecular probes for in vivo PET/SPECT imaging. The scaffolds from which these newer radiolabeled probes are derived include chalcone (5) and its conformationally restricted analogues flavone (6) and aurone (7); stilbene (8) and its analogues diphenyl-1,2,4-oxadiazole (9) and diphenyl-1,3,4-oxadiazole (10); and thioflavin-T analogues such as benzothiazole (11), benzoxazole (12), benzofuran (13), imidazopyridine (14), and benzimidazole (15); as well as quinoline (16) and naphthalene (17) derivatives (Figure 1B). In this review, we provide an overview of these AD radiolabeled early-diagnostic probes according to their scaffolds, with a special emphasis on their synthesis as well as their structure–activity and brain-kinetics relationships. We also provide a brief summary of the latest developments related to the detection of Aβ plaques by near-infrared fluorescence (NIRF) imaging.

Review

Radiolabels used in PET/SPECT molecular probes

Even though they decay rapidly, [¹¹C] (t_1/2 = 20 min) and [¹⁸F] (t_1/2 = 110 min) are the most commonly used radiolabels in PET/SPECT molecular probes for in vivo imaging of Aβ plaques [4]. With a half-life (t_1/2) of 6.01 h compatible with the localization and residence time necessary for imaging, technetium-99m [^99mTc] is also a radionuclide of choice that is easily produced by a ⁹⁹Mo/^99mTc generator [12]. Iodine isotopes such as [¹²⁵I] are also employed, although much less frequently [13]. The general synthetic methods utilized to introduce radiolabels into PET probes are outlined in Scheme 1. These general strategies will be abbreviated as Gs A–D in all subsequent schemes in this review.

Chalcone and its conformationally restricted analogues

Chalcone derivatives

Chalcones and indolochalcones, such as 18a–l, 19a,b, 20a,b, and 21, have been widely reported as Aβ-imaging tracers (Scheme 2A). Structure–activity-relationship (SAR) studies on fluorinated chalcones 18a–l have shown that, in general, chalcones with tertiary amines in their structures demonstrate good affinity for Aβ plaques in in vitro models (K_i = 20–50 nM) (Table 1) [14]. Dimethylation of the amino group seems to be crucial for Aβ binding, since analogues with free amino groups or monomethyl amino groups revealed lower affinity [14]. On the other hand, pegylation is not that essential for plaque binding as tertiary amine analogues with different degrees of pegylation (n = 0–3) all showed similar affinity. In biodistribution experiments using normal mice, the [¹⁸F]-labeled chalcone 19a showed high brain uptake rate and good clearance, whereas the [¹¹C]-labeled chalcone 19b revealed reasonable brain uptake rate, but very fast clearance [14]. The [¹⁸F]-labeled and [¹¹C]-labeled chalcones 19a and 19b were synthesized using similar methods, and a representative synthesis of 19a is shown (Scheme 2B). Aldol condensation between the appropriate acetophenone 22 and benzaldehyde 23 afforded the chalcone backbone, which was subsequently pegylated to give 24 and radiolabeled to give 19a. Compound 19b was generated by using p-nitrobenzaldehyde instead of the corresponding dimethylamine 23 used in the preparation of 19a [14]. The resultant nitrochalcone was then reduced by SnCl₂ in EtOH to yield the free amine, which was monomethylated by controlled addition of an equimolar amount of MeI. The final [¹¹C]-labeled compound 19b was produced by reacting [¹¹C]CH₃OTf with the secondary amine precursor.

[Re]- and [^99mTc]-labeled chalcone analogues 20a and 20b were also studied (Scheme 2A) [12]. The [Re]-labeled analogue 20a displayed higher affinity for Aβ plaque than did the corresponding [^99mTc]-derived compound 20b. However, 20b showed better brain pharmacokinetics than 20a, as indicated by its high brain-uptake rate (1.48% ID/g) and rapid wash out from the CNS (0.17% ID/g at 60 min). Compounds 20a and 20b were synthesized by reacting a Boc-protected metal chelator (Scheme 1C) with 4-O-(bromopropyl)hydroxychalcone 25 (Scheme 2C). After removal of the Boc protecting group, the final [Re]- and [^99mTc]-labeled chalcones 20a and 20b were obtained by treatment with (PPh₃)₂ReOCl₃ and ^99mTcGH, respectively [12].

Finally, the radioiodinated indolochalcone 21, among a series of other derivatives, was prepared through condensation of 4-iodoacetophenone (26) and indole-5-carboxaldehyde (27) to give 28, which was radiolabeled to give the target compound (Scheme 2D) [15]. The indolochalcone 21 showed good binding affinity for Aβ_1-42 aggregates with a K_i < 10 nM. Replacement of the iodo substituent with a chloro, bromo, methoxy, or dimethylamino substituent all gave similar results, but replacement with a fluoro, hydroxy, amino, or methylamino substituent all reduced affinity to varying degrees. Autoradiography in sections of brain tissue from an AD animal model showed that 21 specifically labeled Aβ plaques, but its efficacy was hampered by low in vivo uptake into the brain (0.41% ID/g at 2 min) [15].

Conformationally restricted chalcones: flavones and aurones

Flavones and aurones, such as 29a–c, 30a,b, 31a–c, 32, and 33a,b (Scheme 3A), can be classified as conformationally restricted chalcone derivatives as their basic structures result from insertion of an oxygen atom between the double bond and the phenyl ring attached to the carbonyl group of the chalcone scaffold (Figure 1B, with oxygen atoms depicted in red). The affinity of flavonoids towards Aβ aggregates was first established by using fluorescence staining in brain sections of Tg2576 transgenic mice [10]. The absence of spots in wild-type mouse brain sections indicated the specificity of flavonoids towards Aβ aggregates in AD mouse models. The [¹⁸F]-labeled pegylated flavones 29a–c showed high affinity towards Aβ aggregates with K_i values ranging between 5.3 nM for 29a and 19.3 nM for 29c (Scheme 3A) [16]. SAR studies suggest that, as with chalcones, the tertiary amine in these flavones was important for binding and tracing Aβ aggregates in mouse models, as they consistently outperformed secondary and primary amine analogues [16]. Also as with chalcones, the degree of pegylation had only minor effects on binding properties. Compounds 29a–c showed uptake rates indicative of high to sufficient levels for brain imaging (2.89–4.17% ID/g at 2 min) and moderate clearance rates [16]. The flavone backbone of 29a–c was built by acylating 2-hydroxy-5-methoxyacetophenone (34) with 4-nitrobenzoyl chloride (35) and subjecting the resulting 2-acyloxyacetophenone (36) to Baker–Venkataraman rearrangement [17] to afford the 1,3-diarylpropane-1,3-dione 37, which was dehydrated with sulfuric acid to give 38. Subsequent nitro reduction, reductive methylation, methyl ether cleavage, and pegylation gave the nonlabeled precursors 39a–c. The [¹⁸F]-label was introduced by using the standard [K/K222]¹⁸F in DMSO and acetonitrile reaction conditions (Scheme 3B) [16]. The [Re]- and [^99mTc]-labeled flavone complexes 30a and 30b were also prepared by using the procedure described for the synthesis of the [Re]- and [^99mTc]-labeled chalcones 20a and 20b (Scheme 2). The [^99mTc]-labeled flavone complex 30a displayed high Aβ plaque affinity but limited brain uptake [18].

Aurone derivatives have been investigated for their Aβ plaque binding affinity [19]. The [¹²⁵I]-labeled methylamine aurone 31a presented great binding affinity to Aβ aggregates (K_i = 1.2 nM), better than all reported flavones to date. It also showed rapid brain uptake rate (3.17% ID/g at 2 min) and rapid clearance (0.24% ID/g at 60 min) [19]. The effect of the tertiary amine in this aurone scaffold was less pronounced than that seen with chalcones or flavones. The dimethylamine analogue of 31a had approximately six times weaker binding affinity, while the free amine analogue showed only two times weaker affinity. To further enhance the Aβ plaque traceability of 31a, its methylamine moiety was replaced with ethylene oxide to provide compound 31b, which exhibited a K_i value of 1.05 nM in an in vitro binding assay [13]. The brain kinetics of 31b (brain uptake = 4.51% ID/g at 2 min and washout = 0.09% ID/g at 60 min) were found to be slightly better than those of 31a [13]. Addition of 2 or 3 ethylene oxide units or replacement with a hydroxy or methoxy group did not significantly improve the plaque binding affinity and modestly affected the brain kinetics. Replacement of the terminal hydroxy group of 31b with a fluorine atom negatively affected the brain uptake property of compound 31c (2.34% ID/g at 2 min) when compared to 31b [20]. However, it did not affect the washout character of the compound. These results were confirmed by preparation and analysis of the [¹⁸F]-labeled compound 32 [20]. In general, the aurone derivatives 31a–c and 32 were built from the reaction of methyl 5-iodosalicilate (40) with ethyl bromoacetate (41) followed by ester hydrolysis and cyclization to afford 5-iodo-3-coumaranone (42), which, after condensation with the proper benzaldehydes, gave the aurone scaffold 43, which could be radiolabeled (Scheme 3C). As for the chalcone and flavone derivatives, [Re]- and [^99mTc]-labeled aurone complexes 33a and 33b were also prepared [18]. The high affinity for Aβ aggregates observed with the [^99mTc]-labeled aurone 33b was hampered by its weak brain penetration, which made it unsuitable for in vivo application [18].

Stilbene and its analogues

Stilbene derivatives

The SARs of stilbene analogues, such as 44a–f, 45, and 46a,b (Scheme 4A), as Aβ plaque tracers have been thoroughly investigated. In general, it was found that an electron-donating group at each end of the stilbene derivative is essential for Aβ plaque binding affinity [21]. Analysis of stilbenes 44a–f shows that a monomethylated or dimethylated amine at one end of the stilbene core leads to strong binding affinity for Aβ_1-40 aggregates, while a free amine or nitro group reduces affinity. The opposite end of the stilbene core can be substituted with a hydroxy or methoxy substituent with little effect on binding affinity (Table 2). Derivative 44d, which showed good affinity towards Aβ aggregates in vitro (K_i = 6.0 ± 1.5 nM), has been radiolabeled to give N-[¹¹C]methylamino-4'-hydroxystilbene ([¹¹C]44d), and this compound shows excellent labeling of Aβ plaques in TgCRDN8 mouse brain sections by in vitro autoradiography [22].

[¹⁸F]-Labeled stilbene derivatives have enhanced brain kinetics rendering them appropriate for clinical use [23-25]. In order to control the lipophilicity and keep the partition coefficient (log P) value between 1 and 3, which reduces brain nonspecific binding and improves signal-to-noise ratio, additional hydroxy or ethylene oxide unit(s) were added [21]. An early fluorinated stilbene was [¹⁸F]FMAPO (45), which demonstrated high binding affinity for Aβ aggregates (K_i = 5.0 ± 1.2 nM) in assays using human AD brain homogenates [26]. Even though addition of the fluoroalkyl side chain moiety had little effect on the binding affinity and the clearance rate, it improved brain kinetics significantly (from 1.15% ID/g at 2 min for [¹¹C]44d [22] to 9.75% ID/g at 2 min for 45 [26]). Florbetaben ([¹⁸F]BAY94-9172, 46a), another member of the stilbene class, showed strong binding affinity for human AD brain homogenates (K_i = 6.7 ± 0.3 nM) and promising pharmacokinetics [21], and this compound has progressed to clinical trials. Compound 46a was tested clinically on 15 AD patients and a similar number of healthy elderly volunteers [4]. Interestingly, all AD patients showed widespread neocortical binding of 46a, which was quantified by using the standardized uptake value ratio (SUVR) technique [4]. This observation was further supported by another study using a wider sample population where AD patients demonstrated significantly higher SUVRs when compared to healthy patients or patients with other neural diseases such as Parkinson’s disease, mild cognitive impairment, frontotemporal lobar degeneration, dementia with Lewy bodies, and vascular dementia [27]. More recent phase 2/3 clinical trials collectively showed that compound 46a displays a high degree of sensitivity and selectivity in discriminating between patients with probable AD and age-matched healthy controls [28].

A pyridine analogue of 46a, florbetapir ([¹⁸F]AV-45, 46b) was also prepared using a tosylate precursor with Sumitomo modules for radiosynthesis [23,29]. Compound 46b displayed strong affinity for Aβ peptides in AD brain homogenates (K_i = 2.87 ± 0.17 nM), excellent pharmacokinetics [30], and an acceptable safety profile that paved the way to its clinical application in brain imaging [31]. A number of 46b/PET studies have been conducted [23,32-39]. Using 46b as an imaging probe, PET indicated that the drug accumulates explicitly in Aβ-deposition-rich cortical regions in AD patients with minimal accumulation observed in healthy volunteers [40].

In general, the stilbene nucleus was built using the Wadsworth–Emmons reaction, and a representative synthesis of stilbene 46a is shown (Scheme 4B). Initial Wadsworth–Emmons reaction between diethyl (4-nitrobenzyl)phosphonate (47) and 4-methoxybenzaldehyde (48) constructed the stilbene core 49. The target compound 46 was formed from a straightforward sequence of nitro reduction, reductive methylation, methyl ether cleavage, pegylation and radiolabeling. Several synthetic procedures have been described for the preparation of 46a and its precursors in an effort to optimize yield [21,41,42]. The best reported yield and purity was obtained by mixing the mesylate precursor with the fluorinating agent in a modified PET-MF-2V-IT-1 synthesizer and by purifying using plus C18 Sep-Pak cartridges [41]. In the preparation of [¹¹C]44d, the [¹¹C]-methylation of 4-amino-4'-hydroxystilbene was carried out using the “LOOP” method, in which trapping and reaction of [¹¹C]CH₃OTf with the appropriate stilbene analogue takes place inside an HPLC sample loop [43].

Diphenyl-1,2,4- and diphenyl-1,3,4-oxadiazoles

The replacement of the stilbene ethylene linker with different heterocycles is a common strategy in medicinal chemistry used to improve the pharmacokinetics and/or pharmacodynamics of stilbenes (Scheme 5A) [44-46]. In the case of Aβ probes, a series of 2,5-diphenyl-1,3,4-oxadiazoles 50a–f and 3,5-diphenyl-1,2,4-oxadiazoles 51a–e have been studied in this respect (Table 3 and Table 4). Among the 2,5-diphenyl-1,3,4-oxadiazoles, the dimethylamine analogue 50a (K_i = 20.1 ± 2.5 nM) and methoxy analogue 50b (K_i = 46.1 ± 12.6 nM) showed the best affinities towards Aβ aggregates, and radiolabeling has been performed for both of these compounds. In biodistribution studies, the dimethylamine analogue [¹²⁵I]50a showed good brain uptake and washout rates. Although methoxy analogue [¹²⁵I]50b showed poorer brain uptake, its washout rate was increased compared to its dimethylamine counterpart [47]. Interestingly, changing the heteroatom order in the central ring from 1,3,4 (50a–f) to 1,2,4 (51a–e) has great effects on both the physical characteristics and pharmacokinetics of the compounds. The 3,5-diphenyl-1,2,4-oxadiazole analogue 51c was more lipophilic than its 1,3,4 counterpart 50a (log P = 3.22 for 51c and 2.43 for 50a) [47]. In general, even though 3,5-diphenyl-1,2,4-oxadiazoles 51a–e show excellent affinity for Aβ aggregates in in vitro binding experiments (K_i = 4.3–47.1 nM), they show poorer brain uptake rates (1.07–2.06% ID/g at 2 min) and slower washout rates (3.29–2.01% ID/g at 60 min) than their 1,3,4 counterparts [48]. These findings, together with the close structural similarities between compounds 50a–f and 51a–e, highlight the importance of lipophilicity as a factor in controlling brain kinetics [47].

Representative syntheses of radioiodinated oxadiazoles 50a and 51c are shown (Scheme 5B and C). The 1,3,4-oxadiazole core of [¹²⁵I]50a was obtained from the reaction between 4-iodobenzhydrazide (52) and 4-dimethylaminobenzaldehyde (23) in the presence of ceric ammonium nitrate (CAN) followed by subsequent radioiodination of compound 53 (Scheme 5B) [47]. The 1,2,4-oxadiazole core of [¹²⁵I]51c was obtained by DCC/HOBt-mediated condensation of 4-bromobenzamidoxime (54) and p-nitrobenzoic acid (55). Subsequent nitro reduction, reductive methylation, and radioiodination gave [¹²⁵I]51c (Scheme 5C) [48].

Thioflavin-T analogues

Benzothiazoles

Of all the amyloid imaging classes, the benzothiazoles, such as 56a–w, 57, 58a,b, and 59–65 (Figure 2), may well be one of the most prolific and well-studied. The amyloid imaging dye thioflavin-T (4, Figure 1) served as the inspiration for this class of radiotracers in which the ionic charge was removed to increase lipophilicity and to enhance in vivo BBB permeability. Overall, this class of compounds shows high affinity for Aβ aggregates with promising in vivo pharmacokinetics.

One of the earliest radiolabeled benzothiazoles, [¹¹C]6-Me-BTA-1 ([¹¹C]56a; note: BTA = 2-(4'-methylaminophenyl)benzothiazole, Figure 2), was prepared by methylation of 4-(6-methyl-2-benzothiazolyl)aniline using [¹¹C]methyl iodide [49]. Compared to 4, 56a showed greatly increased lipophilicity and improved binding affinity for Aβ_1-40 (K_i = 890 nM for 4 and K_i = 20.2 for 56a). In postmortem AD brain sections, [¹¹C]56a was able to stain both Aβ plaques and neurofibrillary tangles (NFTs), while pharmacokinetic studies in normal mice showed high brain uptake (7.61% ID/g at 2 min) and good washout (2.76% ID/g at 30 min). Additional modification of this scaffold by removal of the 6-Me group gave [¹¹C]BTA-1 ([¹¹C]56b) [50]. Compound 56b was prepared by coupling of p-nitrobenzoyl chloride (35) and 2-aminothiophenol (66) followed by nitro reduction to 67 and methylation using [¹¹C]methyl iodide (Scheme 6A). While showing a near equal binding affinity for Aβ, the decreased lipophilicity of [¹¹C]56b to the ideal level led to improved pharmacokinetics over [¹¹C]56a as evidenced by improved uptake and washout rates in normal mice (12.9% ID/g at 2 min and 1.7% ID/g at 30 min). Compound [¹¹C]56b showed in vivo specificity for Aβ in the brains of PS1/APP transgenic mice, and it was subsequently shown to bind specifically to amyloid deposits in human AD brain homogenates [51].

The addition of a hydroxy group at the 6-position of [¹¹C]56b gave [¹¹C]6-OH-BTA-1 ([¹¹C]56c) [52]. Compound [¹¹C]56c was synthesized by first coupling p-anisidine (68) with p-nitrobenzoyl chloride (35) to give the amide 69, which was subsequently converted to the thioamide by using Lawesson’s reagent and cyclized to form the benzothiazole core 70 (Scheme 6B). Demethylation with BBr₃ and protection of the resulting hydroxy moiety as the methoxymethyl (MOM) ether gave 71. Reduction of the nitro group to 72, methylation using [¹¹C]methyl iodide, and cleavage of the MOM ether gave [¹¹C]56c. Compound 56c showed high affinity for Aβ_1-40 (K_i = 4.3 nM) (Table 5). This synthesis has since been refined to improve radiochemical yields and eliminate the need for a protecting group by use of [¹¹C]CH₃OTf as the methylating agent [53].

The 6-OH group of 56c made it less lipophilic than both 56a and 56b and likely contributed to its moderate brain entry (0.21% ID-kg/g at 2 min) but good clearance (0.018% ID-kg/g at 30 min) in normal mice. Interesting SAR findings on this scaffold from comparison of 56c–t (Table 5) included that the more lipophilic secondary and tertiary amines at the 4'-position were more potent (K_i) than primary amines. Also, in general, substitution at the 6-position seemed to have only a small effect in terms of K_i as 6-OH, -OCH₃, -CN, and -Br gave similar results. However, substitution at the 6-position had a larger effect on pharmacokinetics in the brain, as 6-OH clearly gave the best results [52]. As one of the most successful radiolabeled Aβ imaging probes to date, [¹¹C]56c has subsequently been named Pittsburgh Compound B (PIB).

Additional studies of [¹¹C]56c in humans have been promising and suggest that PET imaging with this compound can provide quantitative information on amyloid deposits in living patients. In postmortem tissue, [¹¹C]56c exhibited specific binding to the amyloid-laden frontal cortex of the AD brain, but little binding to the frontal cortex of the cognitively normal age-matched control brain. Compound [¹¹C]56c also displayed a rapid entry and clearance in the brain of healthy controls, but a marked retention in AD patients in areas of the brain known to contain large amyloid deposits [54]. Additional data suggested that [¹¹C]56c was suitable for early detection of pathological changes in AD patients before a significant loss of cognitive function is apparent [55].

The impact of changing the position of the hydroxy group of 56c was investigated by synthesizing the 4-OH, 5-OH, and 7-OH analogues 56u–w using methods similar to those described above [56]. The K_i values for these analogues in human AD brain homogenates were between 11–19 nM, indicating slightly reduced affinity compared to 56c (K_i = 2.8 nM) (Table 6). However, each radiolabled analogue was able to stain plaques in sections from transgenic AD mouse brain and human AD brain. The 5-OH analogue [¹¹C]56v showed the best pharmacokinetic profile in normal mice with high brain uptake and a washout rate, that was 8 times faster than that of [¹¹C]56c. Interestingly, it was noted that the 4-OH analogue 56u could form an intramolecular hydrogen bond (i.e. an extra pseudo ring), which could act to increase the lipophilicity of the compound and lead to nonspecific binding and residual background activity in the brain.

A [³H]-labeled analogue of 56c, AZD2184 (57), was also synthesized to give a higher signal-to-background ratio by virtue of its decreased lipophilicity [57]. This compound was prepared through palladium catalyzed Suzuki coupling of the starting halide 73 and boronic acid 74 followed by N-methylation of 75 with [³H]methyl iodide and O-demethylation with sodium thiophenoxide (Scheme 6C). Compound 57 showed high affinity for Aβ_1-40 fibrils in vitro (K_d = 8.4 nM) and lower background binding levels than 56c. While 57 was able to label amyloid deposits in APP/PS1 mice, its brain penetration was not as high as that of [¹¹C]56c.

Besides [¹¹C], other radiolabels have been investigated for benzothiazole imaging agents. Two of the earliest [¹²⁵I]-labeled imaging agents reported were [¹²⁵I]TZDM (58a) and [¹²⁵I]TZPI (58b) [58]. The synthesis of these agents was achieved in two steps by condensation of 5-bromo-2-aminobenzenethiol (76) and the appropriate benzaldehyde 77 followed by radiolabeling of 78 (Scheme 6D). Both 58a and 58b showed high affinity for Aβ_1-40 and Aβ_1-42 aggregates with K_d values ≤0.15 nM in all cases. However, pharmacokinetics for these agents were less than ideal as both showed long retention in the brains of normal mice, which is indicative of nonspecific binding.

A series of iodinated benzothiazoles 59a–o was synthesized using methods similar to those described above and SAR studies were performed (Table 7) [59]. Among the interesting findings was that the introduction of 3'-iodo increased lipophilicity and binding to Aβ_1-40 when R² = NHMe. However, the opposite effect on binding was observed when R² = OH. Among the [¹²⁵I]-labeled derivatives, more polar compounds exhibited better clearance and less nonspecific binding in the brains of normal mice, a typical result for brain imaging probes. One of the most promising compounds identified in this study was [¹²⁵I]59d.

The [¹²⁵I]-labeled benzothiazole bithiophene 60 was synthesized by condensation of 5-bromo-2-aminobenzenethiol and 2,2'-bithiophene-5-carbaldehyde followed by installation of the radiolabel. In in vitro binding experiments, 60 displayed high affinity for both Aβ_1-40 and Aβ_1-42 aggregates with K_i values of 0.25 nM and 0.31 nM, respectively. In addition, it was used to clearly visualize Aβ plaques in AD brain sections and showed favorable pharmacokinetics in the brain with high uptake (3.42% ID/g at 2 min) and fast washout (0.53% ID/g at 60 min).

The [¹²⁵I]-labeled phenyldiazenyl benzothiazole 61 was prepared via a diazo coupling reaction between 79 and 80 to give 81 followed by installation of the radiolabel (Scheme 6E) [60]. Interestingly, in in vitro binding experiments, 61 displayed higher affinity for tau aggregates (K_i = 0.48 nM) than for Aβ aggregates (K_i = 8.24 nM). Although it was used to clearly visualize NFTs in AD brain sections, further modifications will be necessary to improve the pharmacokinetics of this compound in the brain, as it showed particularly slow washout rate (2.89% ID/g at 60 min).

Three [¹⁸F]-labeled analogues of 56c, [¹⁸F]O-FEt-PIB (62), [¹⁸F]FBTA (63), and [¹⁸F]3'-F-PIB ([¹⁸F]GE067, 64) were also prepared. Compound 62 was synthesized by using the hydroxy group of 56c to displace the tosylate of [¹⁸F]fluoroethyltosylate. Compound 62 had a K_i value of 0.17 nM for AD brain homogenate and was able to stain Aβ plaques in postmortem AD brain [61]. Although its biodistribution was not as good as that of 56c, 62 still showed promise in an in vivo study using a rat model of AD [62]. Moving the [¹⁸F]fluoroethoxy substituent of 62 from the 6-position to the 3'-position resulted in a low binding affinity for Aβ and an inability to stain plaques in postmortem AD brain [63]. In compound 63, the [¹¹C]methylamino group of [¹¹C]56c was replaced by a [¹⁸F]fluoroethylamino group, and, while this compound showed better binding affinity than 56c, its brain pharmacokinetics were not as good [64]. Compound 64 also showed promising results in whole-body biodistribution and radiation dosimetry studies [65].

A series of fluorinated benzothiazoles 65a–e was synthesized by direct substitution of the nitro group of a key synthetic intermediate 82 (prepared using synthetic steps already describe for [¹¹C]56c) by an [¹⁸F] atom (Scheme 6F) [66,67]. Compounds 65a,b,d (R = H, Me, and OMe) all showed high binding affinity for AD brain homogenates with K_i values below 10 nM, which is comparable to that of 56c in the same assay, while 65c (R = OH) showed slightly reduced affinity (Table 8). In addition to showing a promising ability to stain Aβ plaques in vivo, [¹⁸F]65a,b showed high brain uptake and rapid washout in normal mice. In fact, each of these compounds displayed better pharmacokinetics than [¹¹C]56c in the same assay.

Benzothiazole probes such as 83a,b and 84a,b labeled with [Re] and [^99mTc] were also synthesized (Scheme 7A) [68,69]. [Re] and [^99mTc]MAMA-BTA (83a and 83b; MAMA = monoamine-monoamide bisthiol-BTA) were prepared by first linking 2-(4-nitrophenyl)-6-hydroxybenzothiazole (85) via 1,5-dibromopentane (86) to monoamine-monoamide bisthiol protected with p-methoxy benzyl (MAMA-PMB, 87) to give 88 (Scheme 7B) [68]. Nitro reduction of 88 followed by thioether deprotection gave MAMA-BTA (89), which was labeled through reaction with the [Re] (used for in vitro studies) or [^99mTc] precursors to give the desired 83a,b. [Re] and [^99mTc]BAT-BTA (84a and 84b; note: BAT = bis(aminoethanethiol)) were prepared by addition of ethyl bromoacetate (41) to the unprotected amine of the S,S'-bis-trityl-N-Boc-1,2-ethylenedicysteamine chelating agent (90) followed by saponification that gave the free acid intermediate 91, which was coupled with 2-(4-aminophenyl)-1,3-benzothiazole (92) (prepared from 2-aminothiophenol (66) and 4-aminobenzoic acid (93)) by using EDCI·HCl and HOBt (Scheme 7C) [69]. Deprotection followed by reaction with the [Re] or [^99mTc] precursors gave 84a and 84b. While both 83a and 84a showed promise as in vitro Aβ labeling agents, the [^99mTc] analogues 83b and 84b exhibited problems in pharmacokinetic studies in vivo. Compound 83b showed sufficient initial uptake (1.34% ID/g at 2 min), but delayed washout (0.65% ID/g at 60 min) in normal mice, while 84b was unable to cross the BBB to a sufficient degree.

Benzoxazoles

Replacement of the sulfur of the benzothiazole backbone by oxygen affords the benzoxazole backbone. Compounds 94, 95a–n, and 96–99 (Figure 3) have also been successfully employed for radioimaging of Aβ plaques. The isosteric replacement of the sulfur of [¹²⁵I]TZDM (58a) with an oxygen was designed to decrease molecular weight and increase lipophilicity and afforded [¹²⁵I]IBOX (94) [70]. Compound 94 was prepared via boric acid catalyzed condensation of 5-nitro-2-aminophenol (100) and 4-dimethylaminobenzoic acid (101) to give the nitro intermediate 102, which was reduced through catalytic hydrogenation to the amine (Scheme 8A). Subsequent conversion to the diazonium ion and displacement with iodide ion gave IBOX, which was radiolabeled to give 94. Compound 94 showed similar affinity for Aβ_1-40 aggregates when compared to 58a, and it was able to label Aβ plaques in postmortem AD brain sections. Importantly, 94 showed superior peak brain uptake (2.08% ID/g at 30 min) and faster brain washout than 58a in normal mice.

Expanding on this 2-arylbenzoxazole scaffold, a series of benzamide-substituted 2-arylbenzoxazoles 95a–n was synthesized [71]. A representative synthesis of [¹²³I]95e is shown (Scheme 8B). Boric acid catalyzed condensation of 4-nitro-2-aminophenol (103) and 4-dimethylaminobenzoic acid (101) gave the nitro intermediate 104. Catalytic hydrogenation as above gave the amine intermediate, and subsequent reaction with 4-iodobenzoyl chloride (105) and installation of the radiolabel gave the target compound. SAR analysis of the compounds indicates that the benzamide moiety is favored at position 5 rather than 6 of the benzoxazole core in terms of binding affinity for Aβ plaques in vitro (Table 9). The best compound was 95e, which had a K_i value of 9.3 nM, but [¹²³I]95e was unable to cross the BBB in vivo. This disappointing result could be appointed to the excessively high lipophilicity of the compound.

To improve the pharmacokinetic profile of 94, the [¹⁸F]-labeled analogue 96 was designed as an imaging probe [72]. Compound 96, which contains an [¹⁸F] end-capped polyethylene glycol chain at position 5 of the benzoxazole core in place of the [¹²⁵I] of 94 at position 6 to reduce lipophilicity, was prepared by polyphosphoric acid catalyzed condensation of 2-amino-4-methoxyphenol (106) and 4-monomethylaminobenzoic acid (107) to give the benzoxazole core, which was O-demethylated to give 108 (Scheme 8C). Subsequent coupling with 2-[2-(2-chloroethoxy)ethoxy]ethanol (109) gave 110. TBDMS protection of the alcohol and Boc protection of the amine gave 111. Finally, TBAF cleavage, installation of the radiolabel, and acid cleavage gave 96. Compound 96 showed good affinity for Aβ_1-42 (K_i = 9.3 nM). This compound also showed promising pharmacokinetics in normal mice, with greatly improved uptake and washout rates compared to 94, and it successfully labeled Aβ plaques in vitro. In addition, it showed increased retention in vivo in transgenic AD mice compared to wild-type. A N,N-dimethyl derivative was also synthesized, and, while it too showed good affinity for Aβ_1-42, its increased lipophilicity compared to the monomethyl compound gave slightly worse pharmacokinetic properties.

The [¹¹C]-labeled styrylbenzoxazole [¹¹C]BF-145 (97) and the related [¹⁸F]-labeled styrylbenzoxazole [¹⁸F]BF-168 (98) were prepared and studied for Aβ imaging [73-75]. The simple two-step synthesis of 97 used polyphosphoric acid trimethylsilyl ester (PPSE) catalyzed condensation of 4-fluoro-2-aminophenol (112) with a cinnamic acid 113 to give the benzoxazole core followed by conversion to the primary amine 114 and radiolabeling (Scheme 8D). The synthesis of 98 was more complex and began with a reaction between 2-methyl-6-methoxybenzoxazole (115) and 4-((N-Boc-N-methyl)amino)benzaldehyde (116) followed by a dehydration reaction to give 117 (Scheme 8E). Subsequent removal of the Boc group followed by installation of a trifluoroacetamide and O-demethylation gave the intermediate 118 used in a Mitsunobu reaction with 2-hydroxyethyl tosylate (119). Amine deprotection to 120 and installation of the [¹⁸F] label gave the target compound 98.

Both 97 and 98 showed good affinity for Aβ_1-42 aggregates (K_i = 4.5 nM and 6.4 nM, respectively). Interestingly, while 98 was able to selectively stain senile plaques (SPs) and NFTs in AD brain sections, 97 was only able to stain SPs. In addition, 97 and 98 showed substantial brain uptake and fast washout (4.4% and 3.9% ID/g at 2 min and 1.6% and 1.6% ID/g at 30 min, respectively) with promising in vivo imaging results in transgenic mice.

Building on the promising results of 97 and 98, an optimized derivative, [¹¹C]BF-227 (99), was studied for Aβ imaging. The key difference in 99 is the replacement of a phenyl ring with a thiazole ring. Compound 99 demonstrated good affinity for synthetic Aβ_1-42 aggregates (K_i = 4.3 nM), rapid uptake (7.9% ID/g at 2 min) and clearance (0.64% ID/g at 60 min) in normal mice, the ability to selectively stain Aβ plaques in AD brain sections, and promising results in a clinical PET study in AD patients [76]. Additional studies suggest that 99 has the possibility to be useful for early detection of AD and also for predicting progression from mild cognitive impairment to AD [77,78]. Interestingly, 99 has also shown promise for diseases other than AD. It has been suggested that 99 may provide a means of diagnosis and disease monitoring in transmissible spongiform encephalopathies [79] and may be useful for monitoring α-synuclein deposits in conditions such as multiple system atrophy and Parkinson’s disease [80]. A version of 99 labeled with [¹⁸F] rather than [¹¹C] has also been proposed for use in Parkinson’s disease [81].

Benzofurans

Replacement of the nitrogen of the benzoxazole backbone with carbon affords the benzofuran backbone of compounds 121–126 (Figure 4), which has also been successfully employed for radioimaging of Aβ plaques. The [¹¹C]-labeled benzofuran 121 was prepared via Wittig reaction between the triphenylphosphonium salt of 2-hydroxy-5-methoxybenzyl alcohol (127) and 4-nitrobenzoyl chloride (35) to give 128 followed by nitro reduction and O-demethylation to give 129 and radiolabeling (Scheme 9A). Using AD brain gray matter homogenates, compound 121 showed good binding affinity for Aβ plaques (K_i = 0.7 nM) and was able to stain both SPs and NFTs in vitro. In normal mice, this compound showed rapid uptake (4.8% ID/g at 2 min) and fast washout (0.2% ID/g at 60 min). In vivo plaque labeling in APP transgenic mice was also successful [82].

Using similar chemistry as described above, a series of iodinated 2-arylbenzofurans 122a–e was prepared and studied [83]. A representative synthesis of [¹²⁵I]122a, which uses a similar Wittig as in the synthesis of 121, is shown in Scheme 9A. It was found that the iodo substituent could be varied between the 5- and 6-positions, and the N,N-dimethylamino substituent could be changed to a secondary methylamino or hydroxy moiety with little effect on binding affinity for synthetic Aβ_1-40, as all compounds had a K_i ≤ 8 nM (Table 10). While the [¹²⁵I]-labeled benzofurans in this series showed good brain uptake in normal mice, their washout was rather slow indicating nonspecific binding in vivo.

Several [¹⁸F]-labeled benzofurans have been employed with success for Aβ imaging. [¹⁸F]FPYBF-1 (123a), which has a N,N-dimethyl-2-aminopyridine group attached to the benzofuran core, was synthesized via Suzuki coupling between 5-methoxybenzofuran-2-boronic acid (130) and 2-amino-5-iodopyridine (131) to give 132, which was followed by reductive amination and O-demethylation to give 133 (Scheme 9B). Finally, reaction with 2-[2-(2-chloroethoxy)ethoxy]ethanol (134) followed by radiolabeling gave 123a. This compound showed high affinity for Aβ_1-42 aggregates (K_i = 0.9 nM), the ability to label plaques in postmortem AD brains, and suitable pharmacokinetic properties in normal mice (5.16% ID/g at 2 min and 2.44% ID/g at 60 min). In addition, it showed good in vivo binding to plaques in transgenic mice [84]. A closely related compound, [¹⁸F]FPHBF-1 (123b), which has a N,N-dimethylaniline group in place of the N,N-dimethylaminopyridine group, was prepared using a Wittig reaction between the triphenylphosphonium salt of 2-hydroxy-5-methoxybenzyl alcohol (127) and 4-dimethylaminobenzoyl chloride (35), followed by O-demethylation and installation of the [¹⁸F]-labeled linker (Scheme 9C). Like 123a, compound 123b showed good affinity for Aβ aggregates in vitro and in vivo. However, its slow washout from the brain, which can be attributed to its increased lipophilicity compared to 123a, indicated that additional refinements will be needed [85].

Derivatives of 123a and 123b were also prepared [86]. These compounds, [¹⁸F]FPYBF-2 (123c) and [¹⁸F]FPHBF-2 (123d), have a secondary methylamino group in place of the dimethylamino group. Introduction of the secondary amine served to reduce lipophilicity. In addition, as the secondary amines are less rapidly metabolized than the tertiary amines, they may help improve the stability of these compounds in vivo. The synthesis of these derivatives used methodology similar to that already described for 123a and 123b. One key difference, however, was the need for orthogonal TBS and Boc protection/deprotection to prevent the secondary amine from reacting with the MsCl used to introduce the radiolabel. Both 123c and 123d showed good affinity for Aβ_1-42 aggregates (K_i = 2.41 nM and 3.85 nM, respectively) as well as the ability to label plaques in transgenic mice. Also, both 123c and 123d showed high uptake and rapid washout with improved pharmacokinetic properties when compared to 123a and 123b.

The [³H]-labeled AZD4694 (124) also showed promise for Aβ imaging [87]. With good affinity for β-amyloid fibrils in vitro (K_d = 2.3 nM), this compound was able to label plaques in human AD brain sections with little nonspecific binding. In addition, the good pharmacokinetic profile of 124 warrants further investigation in vivo.

[Re] and [^99mTc]-labeled benzofurans, BAT-Bp-2 (125a,b), were synthesized from 132 by reductive monoamination and O-demethylation to give 135 (Scheme 9D) [88]. Subsequent reaction with the protected chelation ligand TRT-Boc-BAT-Br (136) and labeling through reaction with the rhenium (used for in vitro studies) and technetium precursors gave compounds 125a and 125b, respectively. Compound 125b showed decent affinity for Aβ_1-42 aggregates (K_i = 32.8 nM) in vitro, although, by comparison to other benzofuran probes of similar structure, it was clear that introduction of the BAT chelator decreased binding affinity. In contrast to other [^99mTc]-labeled Aβ probes, 125b showed decent brain uptake and washout rates in normal mice (1.80% ID/g at 2 min and 0.79% ID/g at 60 min). In addition, it was able to label Aβ plaques in vivo in transgenic mice, a first for a [^99mTc]-labeled Aβ probe.

The [¹²⁵I]-labeled probe 126 contains the benzofuran core, but could also be classified as a chalcone, specifically a chalcone in which the conformation around the double bond is fixed [89]. Compound 126 was synthesized by using a Rap–Stoermer condensation between the bromo-substituted salicylaldehyde (137) and α-brominated 4-nitroacetophenone (138) to form the benzofuran core (Scheme 9E). Nitro reduction followed by methylation gave 139, and radiolabeling gave 126. This compound showed good affinity for Aβ_1-42 aggregates (K_i = 6.6 nM). Secondary methylamino and primary amino derivatives showed decreased binding affinity and poorer labeling of plaques in brain sections from transgenic mice. While the pharmacokinetics of this compound in normal mice were promising (3.53% ID/g at 2 min and 0.87% ID/g at 60 min), they were not as good as those previously reported for [¹²⁵I]-labeled N,N-dimethylamino chalcones and aurones.

Imidazopyridines

The imidazopyridine core has also been used in developing novel Aβ imaging agents such as 140–142 (Scheme 10A). Initial SAR studies were based on derivatives 140a–j. One of the most successful imidazopyridines studied to date has been [¹²⁵I]IMPY ([¹²⁵I]140e). Representative of this scaffold, the synthesis of [¹²⁵I]140e used a fusion reaction between 2-amino-5-iodopyridine (143) and an α-bromoacetophenone 144 to form 140e, which was then radiolabeled (Scheme 10B) [90]. This preparation has since been improved by Kung et al. who, through the use of a reverse-phase C4 minicolumn with stepwise washing and elution, have simplified the purification process by eliminating the need for HPLC purification [91]. While 140e showed good affinity for Aβ_1-40 aggregates (K_i = 15.0 nM), SAR analysis demonstrated that, in general, other modifications to the scaffold were not well tolerated and reduced the binding affinity (Table 11). An exception was the replacement of the 6-iodo substituent with a bromine, as 140f showed similar affinity to the parent compound [90].

In addition to its good binding affinity for Aβ, [¹²⁵I]140e showed other promising properties for use as an imaging probe. For example, it selectively labeled plaques in postmortem AD brain sections and showed plaque labeling with low background activity in a transgenic mouse model. The pharmacokinetics of [¹²⁵I]140e were also promising. It showed high uptake (2.9% ID/g at 2 min) and fast washout (0.2% ID/g at 60 min) in normal mice. These kinetic properties represented improvements over both 58a and 94 [92]. Safety, biodistribution, and dosimetry studies of [¹²³I]IMPY, the [¹²³I]-labeled counterpart of 140e, have indicated it may be a safe radiotracer with appropriate pharmacokinetics for use in AD patients [93].

More in depth SAR studies of the imidazopyridine scaffold have been conducted by synthesizing the series of analogues 141a–w [94]. The effect of different substituents on binding affinity (K_i) for human Aβ plaques was examined (Table 12). In general, it was found that the N,N-dimethylamino analogues (R¹ = R² = Me) had higher binding affinity for human Aβ plaques than did the secondary methylamino analogues (R¹ = H, R² = Me). Little tolerance for substitution at both R³ and R⁵ was seen, as the most potent compounds almost always had a hydrogen atom at these positions. One exception was observed with the secondary methylamino analogue when R³ = Br, as 141n showed high affinity. For R⁴ it was seen that polarizable or electron-withdrawing substituents showed higher affinity than strongly electron-donating substituents. In addition, it was observed that bulky, hydrophobic thioether substituents (such as R⁴ = SCH₂C₆H₄-p-OMe) were well tolerated at this position. This finding was of particular interest as it provided a possible means of generating new PET ligands via [¹¹C]- or [¹⁸F]-labeling through S-alkylation.

The [¹⁸F]-labeled imidazopyridine, [¹⁸F]FPPIP (142), was prepared starting from 140e. A palladium-catalyzed coupling with tributyl(vinyl)tin to give an alkene intermediate was followed by hydroboration-oxidation to give the hydroxypropyl intermediate 145, which was radiolabeled to give 142 (Scheme 10B). This compound showed good binding affinity for Aβ (K_i = 48.3 nM) in using human AD cortical tissues, as well as specific labeling of Aβ plaques in postmortem AD brain. This, coupled with favorable pharmacokinetics observed in a normal rhesus monkey, made 142 a promising compound [95]. However, another [¹⁸F]-labeled imidazopyridine, [¹⁸F]FPM-IMPY, has shown less promising results. This compound, in which one of the N-methyl groups of IMPY was replaced with a [¹⁸F]fluoropropyl moiety, showed lower binding affinity than IMPY and poor pharmacokinetics [96].

Benzimidazoles

The benzimidazole scaffold is highly similar to the imidazopyridine scaffold, but the benzimidazole ring has reduced lipophilicity when compared to imidazopyridines. This has the potential to reduce nonspecific binding and enhance signal-to-noise ratio. The [¹²⁵I]-labeled benzimidazole analogue of 140e, compound 146, was prepared through cyclization of 4-bromobenzene-1,2-diamine (147) and p-dimethylaminobenzaldehyde (23) to give 148 followed by installation of the radiolabel (Scheme 11) [97]. Compound 146 showed good binding affinity for Aβ_1-42 aggregates (K_i = 9.8 nM), as well as high uptake and rapid clearance in normal mice (4.14% ID/g at 2 min and 0.15% ID/g at 60 min). In vitro labeling of Aβ plaques in AD brain sections showed a strong signal with low background, and in vivo plaque labeling in transgenic mice was also successful. However, this scaffold is lacking in detailed SAR analysis compared to the imidazopyridine scaffold.

Quinoline and naphthalene analogues

Quinolines

Investigation of the quinoline scaffold for imaging in AD has yielded some interesting results, despite there only being a few examples in the literature. The [¹⁸F]-labeled 2-fluoroquinolin-8-ol [¹⁸F]CABS13 (149) has recently been reported (Figure 5) [98]. The straightforward synthesis of this compound began with benzyl protection of 2-chloroquinolin-8-ol followed by installation of the [¹⁸F]-label and Pd-catalyzed hydrogenolysis to give the target compound. Compound 149 potently bound to Aβ-Zn aggregates (K_d = 1.5 nM) and showed rapid uptake and washout in normal mice (10% ID/g at 2 min and 1.1% ID/g at 30 min). Also, delayed washout of 149 was observed in APP/Ps1 transgenic mice, which was indicative of non-specific binding to Aβ plaques. However, two other quinoline probes, [¹¹C]BF-158 (150) and [¹⁸F]THK523 (151), had high affinity for tau pathology as opposed to Aβ. Compound 150 showed good uptake and washout in normal mice (11.3% ID/g at 2 min and 2.1% ID/g at 60 min), and was able to label NFTs in postmortem AD brain section while only faintly staining plaques [99]. Compound 151 showed high affinity (K_d = 1.7 nM) and selectivity for recombinant tau fibrils in vitro, and, with favorable pharmacokinetics, it was able to highlight tau pathology in vivo in transgenic mice [100].

Naphthalenes

Replacement of the cyclic nitrogen in the quinoline scaffold described in the previous section affords the naphthalene scaffold. This scaffold has shown promising results for Aβ imaging, particularly [¹⁸F]FDDNP (152), although this scaffold is also represented by only a few examples in the literature. Compound 152 was prepared starting from 1-(6-hydroxy-2-naphthyl)-1-ethanone (153) via a Bucherer reaction with 2-(methylamino)ethanol (154) followed by Knoevenagel reaction of 155 with malononitrile (156) and [¹⁸F] labeling of 157 (Scheme 12) [101,102]. Compound 152 bound to synthetic Aβ_1-40 fibrils with high affinity (K_d = 0.12 nM) and crossed the BBB [103]. In addition, PET imaging studies using 152 demonstrated the ability of this compound to determine the localization and load of both SPs and NFTs in living AD patients [104], as well as the ability to differentiate between patients with no cognitive impairment, mild cognitive impairment, and AD [105].

[Re]- and [^99mTc]-labeled derivatives of 152 have also been prepared by bromination of 157 with NBS to give 158 followed by conjugation with MAMA-PMB (87) and deprotection with acid to give 159. Reaction with the technetium or rhenium precursors gave the target derivatives 160a,b (Scheme 12) [106]. In vitro binding studies with the [Re]-labeled compound 160a showed a 14-fold decrease in binding affinity for Aβ_1-42 aggregates compared to 152. In addition, the [^99mTc]-labeled compound 160b showed very low brain uptake in normal mice indicating the need for additional refinements of this compound.

Combination of known scaffolds

With the success of the benzothiazole and imidazopyridine scaffolds for Aβ imaging, it was logical to suspect that combination of the two scaffolds into a single molecule could also provide a good imaging agent. Examples of this combination scaffold can be seen in compounds 161–164 (Scheme 13A). IBT (161) was prepared by direct coupling of 6-methoxybenzo[d]thiazol-2-amine (165) and the nitro substituted α-bromoacetophenone 138 to give 166 followed by installation of the radiolabel (Scheme 13B) [107]. Compound 161 showed good affinity for both Aβ_1-40 and Aβ_1-42 (K_i = 3.5 nM and 5.8 nM, respectively) and was comparable to compound 56c in the same assay. The pharmacokinetics of this compound were also similar to those of [¹¹C]56c. In vivo specific plaque labeling by compound 161 was confirmed through studies in APP/Ps1 transgenic mice. Derivatives of this combination scaffold 162a–n were also investigated (Table 13) [108]. Of note was derivative 162i in which the secondary methylamino group of 161 has been replaced with iodine. This derivative showed high affinity for Aβ_1-40 (K_d = 10.9 nM), and the iodo substituent could readily be radiolabeled. However, the high lipophilicity of this compound may lead to nonspecific plaque labeling in vivo.

The [¹²⁵I]-labeled styrylindole 163 and styrylquinoline 164 scaffolds synthesized by Yang et al. can be thought of as stilbene combination scaffolds [109]. The synthesis of the [¹²⁵I]-labeled styrylindole 163 used a Wittig reaction between the triphenyl phosphonium ylide 167 and 1H-indole-5-carbaldehyde (27) to give 168 followed by radiolabeling (Scheme 13C). The [¹²⁵I]-labeled styrylquinoline 164 was prepared by using an identical synthesis with substitution of the indole by quinoline-6-carbaldehyde. Both 163 and 164 showed good affinity for Aβ_1-40 aggregates (K_i = 4.1 nM and 8.6 nM, respectively). Compound 163 was able to stain Aβ plaques in in vitro brain sections from APP/Ps1 transgenic mice and showed high uptake and rapid clearance in normal mice (4.27% ID/g at 2 min and 0.28% ID/g at 60 min). However, compound 164 showed relatively low brain uptake and slow washout by comparison.

Others

Several other less common scaffolds have been evaluated as Aβ-imaging agents. The [¹²⁵I]-labeled N-methyl-4-anilinophthalimide derivative 169 was prepared and evaluated as a potential probe for Aβ plaques (Scheme 14A) [110]. This compound was generated via a Cu powder-catalyzed coupling reaction between N-methyl-4-aminophthalimide (170) and 1-bromo-4-iodobenzene (171) to give 172, which was then radiolabeled. Compound 169 showed high binding affinity to AD brain homogenates (K_d = 0.21 nM) as well as excellent brain uptake (5.16% ID/g at 2 min) and fast washout (0.59% ID/g at 60 min). SAR studies with other N-methyl-4-anilinophthalimide derivatives demonstrated that a hydrophobic substituent at the 4-position of the aniline ring is important for the binding affinity of this family of compounds.

The [¹²⁵I]-labeled quinoxaline derivative 173 was also synthesized and evaluated for in vivo imaging of Aβ plaques (Scheme 14B) [111]. The quinoxaline backbone of this compound was prepared from the reaction of α-bromoacetophenone 144 and 4-bromobenzene-1,2-diamine (147) in DMSO in a one-pot tandem oxide condensation procedure. This reaction gave the desired 2-aryl-6-substituted quinoxaline 174 as the major product and the isomeric 2-aryl-7-substituted quinoxaline (not shown) as a minor product. The radioiodinated probe was prepared from 174. Compound 173 showed excellent affinity for Aβ_1-42 aggregates in vitro (K_i = 4.1 nM). In addition to being able to specifically label plaques in brain sections from AD patients, 173 readily crossed the BBB showing high uptake into the brain (6.03% ID/g at 2 min). However, with moderate washout (1.12% ID/g at 120 min), additional refinements will be needed to improve the pharmacokinetics of these molecules.

The boron [¹²⁵I]-labeled dipyrro-methane (BODIPY) analogue 175 was prepared to serve as a dual functional SPECT/fluorescence probe for imaging Aβ (Scheme 14C) [112]. Compound 175 was synthesized through Suzuki coupling of the starting boronic acid 176 with 1-bromo-4-iodobenzene (171). Aldehyde reduction of 177 followed by reaction with triphenylphosphine gave the Wittig reagent 178 for reaction with 2-formylpyrrole (179). The Wittig product 180 was condensed with 3,5-dimethylpyrrole-2-carboxaldehyde (181) to form the BODIPY backbone 182. Subsequent installation of the radiolabel gave 175. Although 175 showed decent affinity for Aβ_1-42 aggregates (K_i = 108 nM) and the ability to label plaques in brain sections from transgenic mice, its in vivo use was limited by extremely low brain uptake, which could be attributed to rapid trapping of the compound in the liver.

Fluorescence probes

Although PET is currently the most promising approach for Aβ plaque detection, this technique has two main limitations: (i) the short half-life of positron-emitting nuclei (t_1/2 = 20 min for [¹¹C] and 110 min for [¹⁸F]) and (ii) the narrow availability of this technology that requires a local cyclotron for generating short-lived positron-emitting radionuclides and a synthetic unit to produce radiolabeled agents [113]. Other imaging technologies have been investigated to overcome these problems. Different fluorescence techniques have been reported [114-116]; however, the near-infrared fluorescence (NIRF) imaging technique is the only one that has an in vivo application. Since normal biological tissues reveal limited photon absorbance in the near-infrared region, NIRF seems to be the second most promising Aβ deposits tracer tool [117]. In the following sections, we will briefly cover the different scaffolds that have been explored as NIRF ligands using in vivo models.

Oxazines

The oxazine dyes 183–186 were investigated as Aβ aggregate target-specific probes in the NIRF imaging technique (Scheme 15A) [113]. The preparation of 183 was accomplished through reaction of 4-methyl-3,4-dihydro-2H-benzo[b][1,4]oxazin-6-ol (187) with p-nitrobenzenediazonium ion (188) to give the key azo intermediate 189, which afforded the desired oxazine dye 183 upon further reaction with 4-methyl-3,4-dihydro-2H-benzo[b][1,4]oxazin-6-ol (187) (Scheme 15B). Compound 183 proved to give a higher fluorescence intensity than other derivatives such as 184–186 [113]. Using APP23 transgenic mice and compound 183, Aβ plaques could be traced quantitatively [113].

Thiobarbitals

The thiopental dimer THK-265 (190) (maximal emission wavelength >650 nm) was discovered as a good NIRF imaging ligand by screening a large pool of dye candidates (Figure 6) [117]. Compound 190 displayed high binding affinity towards Aβ aggregates (K_d = 97 nM) [117]. Its usefulness in AD diagnosis was confirmed in an animal model as it provided good discrimination between amyloid deposits in the brain and other normal tissues [117]. Compound 190 was also used in a quantitative correlation of different Aβ aggregation levels with NIRF signals [118].

Quinoxalines

The use of the radiolabeled quinoxalines for imaging Aβ was discussed in the section “Others” of this review. A quinoxaline derivative, compound 191, was also synthesized and explored for fluorescence imaging (Scheme 16) [119]. The synthesis of 191 began with conversion of the starting lactam 192 to the corresponding chloride by using phosphorus oxychloride followed by reaction with hydrazine. Condensation of the resulting hydrazino-derivative 193 with quinoline-4-carboxaldehyde (194) gave 191. Although 191 has not been tested in vivo yet, this compound warrants further investigation as it has shown the ability to selectively stain amyloid structures in brain sections of transgenic mice, as well as the ability to cross the BBB. In addition, the 7-fluoro substituent of compound 191 could potentially be radiolabeled for in vivo application.

Conclusion

In summary, this review covered the main scaffolds used for radioimaging of Aβ plaques, one of the major pathological hallmarks of AD. Highlighted were important synthetic steps for scaffold formation and introduction of radiolabels, SAR findings where appropriate, and binding affinities and brain kinetics of each scaffold. The synthesis of each scaffold presented was fairly straightforward using well-established reactions, and synthetic complexity will likely not impede future development of Aβ chemical probes.

Most of the scaffolds present compounds with good binding affinity for Aβ in vitro. SARs tended to vary between scaffolds so it is impossible, without extensive computational work, to declare certain functionalities necessary for this class of chemical probes. However, the dimethylamine structural feature appears in a number of the compounds with high binding affinity, and it is likely that this functionality is important for Aβ interaction. Representative examples of this can be seen in examining SAR trends for chalcones 18a–l, benzothiazoles 56c–t, or imidazopyridines 141a–w among others. With regards to pharmacokinetics in vivo, results varied between scaffolds and radiolabels. In general, [^99mTc]-labeled compounds showed poor pharmacokinetic profiles with only 125b being able to label plaques in animal studies. A balance in lipophilicity appeared to be particularly important in terms of pharmacokinetics, as imaging probes need to be lipophilic enough to easily penetrate the BBB, but not too lipophilic to avoid nonspecific binding in the brain.

Many of the molecules described in this review showed very promising results for the in vivo imaging of Aβ plaques in humans. For example, stilbenes 46a and 46b, benzothiazole [¹¹C]56c, and naphthalene 152 have been studied in clinical trials with favorable results. The half-life of the radiolabel and overall lipophilicity will continue to be two of the biggest factors for the clinical success of these probes. Future development and testing of these molecules will be of critical importance as the development of Aβ imaging probes will provide an effective means of monitoring new treatments for AD. While not the focus of this review, it should be noted that the current treatments for AD only treat cognitive symptoms and have little to no effect on slowing or reversing the progression of the disease. Current research efforts aimed at developing molecules that target Aβ plaques, specifically inhibition of plaque formation and disaggregation of already formed plaques, could lead to new therapeutics capable of reversing the progression of AD. Probes such as those described herein will play an important role in evaluating the effectiveness of such drugs. Additionally, the development of Aβ imaging probes will likely lead to better and earlier diagnosis of AD, which in turn will allow future treatments to be more effective.